svn commit: r353103 - head/sys/net

[Kyle Evans]

On Fri, Oct 4, 2019 at 2:12 PM John Baldwin <jhb at freebsd.org> wrote:
>
> On 10/4/19 6:43 AM, Kyle Evans wrote:
> > Author: kevans
> > Date: Fri Oct  4 13:43:07 2019
> > New Revision: 353103
> > URL: https://svnweb.freebsd.org/changeset/base/353103
> >
> > Log:
> >   tuntap(4): loosen up tunclose restrictions
> >
> >   Realistically, this cannot work. We don't allow the tun to be opened twice,
> >   so it must be done via fd passing, fork, dup, some mechanism like these.
> >   Applications demonstrably do not enforce strict ordering when they're
> >   handing off tun devices, so the parent closing before the child will easily
> >   leave the tun/tap device in a bad state where it can't be destroyed and a
> >   confused user because they did nothing wrong.
> >
> >   Concede that we can't leave the tun/tap device in this kind of state because
> >   of software not playing the TUNSIFPID game, but it is still good to find and
> >   fix this kind of thing to keep ifconfig(8) up-to-date and help ensure good
> >   discipline in tun handling.
>
> Why are you using d_close for last close anyway?  It's not really reliable compared
> to using cdevpriv and a cdevpriv dtor.
>

This decision predates me by a long time, I'm afraid. =-)

If you have time to elaborate on the comparable reliability point, I'd
be interested in hearing it. A little bit of searching didn't seem to
turn up much there, I'm afraid.

I did otherwise spend a little bit of time diving into the path taken
to get to d_close and the trade-offs between cdevpriv vs. what tuntap
does now. I think I'm convinced either way that cdevpriv is a good way
to go- it seems to have the advantage that with a little refactoring
we could actually set the softc atomically on the device cdevpriv
instead of cdev->si_drv1 and I can axe this rwatson@ comment about the
non-atomic test and set.

I don't see any downside here.